Asking The Wrong Questions

When the subject of AI comes up, the first question most people ask is: “Will I still have a job?” In businesses, it's mostly the same question. “Can we make our teams more efficient and save money?”

This thinking is understandable. But these are the wrong questions.

People's anxiety is legitimate and there will be disruption. AI, in some way, will change every job and some roles will be eliminated. This is what happens when a new general purpose technology reshapes an economy. Companies will find efficiencies and some will reduce headcount. These outcomes are real and they deserve society's serious attention.

But efficiency is the small prize. The leaders and organizations that stop at “how do we do the same work with fewer people” are settling for a fraction of what this moment offers and the individual who thinks of AI only as a threat to defend against will find themselves unprepared for the future.

The more important questions sound different. Why do I do this work in the first place? What do our customers and stakeholders actually need from us? How does the entire process of how we deliver value need to change? And what skills will be in demand as we make that transition?

These are not abstract questions. They are the questions that will separate the organizations that thrive from the ones that automate their way into irrelevance. They are also the questions that will help individuals find their footing in a labor market that is being restructured, not reduced.

Think of all the products we wish we had the time or the money or the technology to create. Big bet opportunities we were too constrained to seize. Think of the value it's now possible for one person to produce. That's the organization we should be and that person should be you.

I am not a futurist and I am not selling a vision of AI utopia. What I see in the current conversation about AI and jobs is a tremendous amount of noise built on the wrong framing. This article is an attempt to cut through that noise.

In “Beyond Transformation,” I argued that layering AI onto broken processes just produces broken outputs faster. But the real opportunity goes further than fixing what is broken. The real opportunity is questioning whether the process should exist in its current form at all.

When electricity first arrived in factories, manufacturers replaced their steam engine with an electric motor. Same layout. Same workflow. Slightly cleaner power source. It took nearly forty years before anyone realized the real advantage: small motors on individual machines could free the factory floor entirely, enabling a new layout organized around the logic of production rather than the logic of power transmission. The productivity gains were enormous, but they required redesigning the system, not just upgrading the engine.¹

Most organizations are in the “replace the steam engine” phase with AI. They are making existing processes faster without questioning the structure of those processes. Sangeet Paul Choudary's book Reshuffle argues that the real impact of AI is not about making tasks more efficient, it is about breaking apart the architecture of how work gets done and reassembling it around new logic. Unbundling what exists and rebundling it into something that was not possible before.² The bottlenecks holding most organizations back are not technical, they are institutional. Our governance models, our workflows, our organizational structures were all designed around constraints that AI is removing. When you remove a constraint that shaped an entire system, the system does not simply speed up. It has to be redesigned.

Consider one of the roles most impacted by AI, software development. At Anthropic, the company that builds Claude, 70 to 90 percent of code is now AI-generated. One engineering lead ships more than 20 code updates per day, all produced by AI. But Anthropic is not shrinking its team. It is hiring more people, shifting toward generalists who specify what needs to be built, evaluate whether it was built correctly, and direct the AI doing the work.³ Morgan Stanley projects the software development market will grow at 20 percent annually, reaching $61 billion by 2029.⁴ The work is not disappearing, it is being reorganized. The bottleneck moved from production to judgment.

This brings us to a pattern that is easy to miss if you are focused only on efficiency. In the 1860s, the economist William Stanley Jevons observed something counterintuitive about coal. As steam engines became more efficient, total coal consumption did not decrease. It dramatically increased. Because cheaper energy made new applications viable that were previously too expensive to consider. This pattern has held for every major efficiency gain in history. Cheaper steel did not reduce the use of steel. It gave us skyscrapers, railroads, and automobiles. Cheaper computing did not reduce the need for computing. It gave us the internet and mobile phones.⁵

This idea, known as Jevons Paradox, applies directly to what AI is doing to human work. AI makes each worker dramatically more productive. But more productive workers does not mean fewer workers are needed. It means that work which was previously too expensive, too slow, or too ambitious to attempt is now within reach. Experiments that were too risky now cost a fraction of what they did. Products that lived on the whiteboard because no one could justify the investment can finally be built. The total volume of valuable work to be done is expanding, not contracting. The pie is growing.

WHOOP, the wearable health company, is nearly doubling its workforce in 2026, adding more than 600 roles while simultaneously investing heavily in AI. Their CEO put it simply: “Companies are debating whether to hire more people or just invest in AI. We are doing both.”⁶

The companies that see AI only as a cost reduction tool will pocket short term savings. The companies that see it as a way to restructure their systems and pursue new ambitions will build the future. And they will need people to do it.

But not the same people doing the same things. The nature of the work is changing. And that requires a new set of skills.

If work is being restructured rather than eliminated, what does that demand from people? I believe it requires three skills in particular: curiosity, specification, and testing.

Curiosity. The people getting the most from AI are learning from it every day. They ask what the frontier is and they push their own skills forward. This is not idle interest. It is curiosity about what AI can do, but equally important, what am I now capable of doing with AI's assistance? Without this drive, you get adequate results. With it, you discover capabilities you did not know existed. The gap between a curious user and a passive one widens every month as the tools improve.

Specification. Requirements, prompt engineering, clear problem definition. These are all the same skill. To get the best output from AI, you must know what the best output looks like. You have to have a destination you are driving to. You can explore and take the scenic route, but if you do not have and can't develop clear specifications, you will not arrive at the right result. This is where domain expertise becomes more valuable, not less. Knowing what to specify requires understanding why the work exists and what value it produces. The person closest to the customer, the process, or the problem is the person best positioned to define what good looks like.

Testing. The final output is the responsibility of the human. AI can produce confident, polished, wrong answers. Testing means being able to identify when the output is correct and helping the agent test and refine its own work. It means not accepting answers without questioning and confirming. You must have a test, even if it is your own informed evaluation, that passes before you allow the work to be finished. The human who cannot evaluate quality is not collaborating with AI. They are being led by it.

A recent study of software developers learning a new programming tool illustrates this clearly.⁷ Developers were split into two groups: one worked with AI assistance, the other without. Those who delegated everything to AI scored dramatically lower on comprehension, averaging below 40 percent. Those who asked conceptual questions, requested explanations alongside generated code, and verified outputs scored above 65 percent. The delegators were not even faster. They gave up learning and got nothing in return.

The high performers in the study were practicing all three skills. They were curious about how the tool worked, not just what it produced. They specified what they needed clearly enough to get useful results. And they tested the output rather than trusting it blindly. The largest gap showed up in debugging, the ability to recognize when code was wrong and understand why. This is exactly the skill that matters most when humans are overseeing AI-generated work.

These are not technical skills reserved for programmers. A nurse who is curious about what AI can do for patient triage, who can clearly specify what a good outcome looks like, and who can evaluate whether the AI's recommendation makes clinical sense is exactly the person who will thrive. A project manager who can articulate clear requirements, question the AI's assumptions, and verify the deliverable against the actual need is more valuable now than they have ever been.

Curiosity, specification, and testing are learnable. Organizations that invest in building these capabilities across their workforce will have a real advantage. The future belongs to people who work with AI as a partner, not people, or organizations who hand their work to AI and walk away.

The question was never “will AI take my job?” The question is how does the entire system of work change, what skills do I need to thrive in it, and how do I work with AI rather than hand my work to it. I believe there will be more meaningful work ahead, not less. More opportunity for people with domain expertise, judgment, and curiosity. But only for those willing to rethink how the work gets done and build toward what comes next.

To be clear, my optimism does not ignore the significant friction felt in transitions. I am talking about massive societal change, and people are not good at change. Many will struggle to make this transition. Some will not be able to make it at all. And as a society, we are doing almost nothing to address the scale of this impact. The same forces that expand opportunity for those who adapt, create real hardship for those who cannot. That is not a footnote. It is the subject of Part 2 of this series.

If you are in a position of leadership, your job is to prepare your people for this change. Invest in curiosity, specification, and testing across your organization. Restructure the work, not just the headcount. Give people a reason to believe this is a future worth building toward.

Start with why. Why does this work exist? Why do our customers need it? Why are we organized this way? The answers will reshape everything that follows.